AI Glossary
The key terms of AI in clear, plain-language definitions — from the basic concept to the technical term.
- Artificial Intelligence (AI)
- The ability of computer systems to perform tasks that normally require human intelligence — understanding language, analysis, and decision-making. A broad term that encompasses machine learning, language models, and agents.
- Machine Learning (ML)
- A branch of AI where systems learn from data instead of being programmed with explicit rules. The more high-quality data, the better the model performs at predicting or classifying.
- Generative AI
- AI systems that produce new content — text, images, code, or audio — rather than only analyzing existing input. Models like Claude and ChatGPT are examples.
- Large Language Model (LLM)
- An AI model trained on vast amounts of text to understand and generate language. It works by predicting the most likely next word, and it is the engine behind chat tools and coding agents.
- Prompt
- The text you write to the model to ask it for something. The quality of the prompt largely determines the quality of the output — crafting it methodically is what we call prompt engineering.
- Read more: Prompt engineering guide
- Prompt Engineering
- The skill of crafting prompts methodically — adding context, a role, examples, and a target format — to get a precise, repeatable result from an AI model.
- Read more: Prompt engineering guide
- AI Agent
- A system built on a language model that doesn’t just answer, but achieves a goal through planning, sequential steps, and using real tools. It completes a task rather than just answering about it.
- Read more: AI agents guide
- Coding Agent
- An AI agent specialized in programming: it reads your project files, writes code, runs it, finds errors and fixes them — a full loop without you intervening at every step. Claude Code is an example.
- Read more: AI agents guide Vibe coding guide
- Vibe Coding
- A way of building software where you describe what you want in natural language, and an AI agent writes, runs, and debugs the code. The term was coined by Andrej Karpathy in 2025.
- Read more: Vibe coding guide
- Token
- The basic unit a model processes text in — it may be a word or part of one. Model usage cost and length limits are counted in tokens, not words.
- Context Window
- The maximum amount of text (in tokens) a model can consider at once, including your question and what preceded it in the conversation. Exceeding it makes the model “forget” the start.
- Hallucination
- When a model produces information that sounds authoritative but is incorrect or fabricated. A core reason why reviewing AI output is essential, especially for figures and sources.
- Retrieval-Augmented Generation (RAG)
- A technique that supplies the model with information retrieved from an external source (documents, a database) at answer time, making its output more current, accurate, and grounded in your data.
- Fine-tuning
- Retraining a ready model on specialized data so it masters a specific task or style. An advanced step; many needs are solved first with prompt engineering or retrieval before resorting to it.
- Embedding
- Converting text (or an image) into a vector of numbers representing its meaning, so similar meanings sit close together. The basis of semantic search, retrieval, and recommendation systems.
- Few-shot Prompting
- Giving the model one or more examples of the desired output inside the prompt so it mimics its shape and tone. With no example it’s zero-shot, with one it’s one-shot — and examples often beat instructions.
- Read more: Prompt engineering guide
- Chain-of-Thought
- A technique where you ask the model to think step by step before the final answer. It reduces errors in tasks needing reasoning or calculation because it shows the logic instead of jumping to a conclusion.
- Read more: Prompt engineering guide
- Multimodal
- A model that understands and processes more than one type of input — text, image, audio — at once, so you can, for example, show it an image and ask about it in text.
- API
- An interface that lets programs talk to each other. Through AI model APIs you build your own applications that call the model programmatically rather than using it only from the chat interface.
- Model Context Protocol (MCP)
- An open standard that connects models and agents to external tools and data sources in a unified way, expanding what an agent can do without building a custom integration for every tool.
- Read more: AI agents guide
Go from term to practice
Our practical guides explain these concepts step by step — by actually building, not theory.
Browse the blog